Changes in Aquatic Insect Emergence in Response to Whole-Lake Experimental Manipulations of Introduced Trout

Freshwater Biology (2009) 54, 982–993 doi:10.1111/j.1365-2427.2008.02145.x

Changes in aquatic insect emergence in response to whole-lake experimental manipulations of introduced trout

KAREN L. POPE*,1, JONAH PIOVIA-SCOTT† AND SHARON P. LAWLER* *Department of Entomology, University of California, Davis, CA, U.S.A. †Center for Population Biology, University of California, Davis, CA, U.S.A.

SUMMARY 1. Insects emerging from mountain lakes provide an important food source for many terrestrial predators. The amount of insect subsidy that emerges from lakes is influenced by predator composition, but predator effects could be ameliorated by greater habitat complexity. We conducted a replicated whole-lake experiment to test the effects of introduced fish predators on the abundance and composition of aquatic insects within and emerging from the littoral zone of 16 mountain lakes in the Trinity Alps Wilderness in northwestern California. 2. Study treatments matched the fisheries management options being implemented in California’s wilderness areas: (i) continued stocking with non-native trout, (ii) suspension of stocking, and (iii) suspension of stocking and removal of fish. We also included four naturally fishless ‘reference’ lakes. We compared abundances and biomass of emerging aquatic insects before treatments were initiated and for 3 years following their establish ment. Abundances of benthic insects were also compared in the third year post-treatment. 3. Trout removal rapidly increased abundances of mayflies, caddisflies, and insect predators, and overall insect biomass emerging from lakes. No significant differences were found between the suspension of stocking lakes and continued stocking lakes. Fish density was a more important predictor of aquatic insect emergence than habitat complexity. 4. Mayfly larvae responded positively to fish removal and caddisfly larvae tended to be more abundant in lakes without fish, but we did not detect effects on abundance of predatory insects. However, we found large insect predators in shallower water in lakes with fish compared to fish removal or fish-free reference lakes. 5. These results provide insights into the continuing effects of past and current fish stocking practices on the flow of insect prey from mountain lakes into the neighbouring terrestrial environment. We also show that these consequences can rapidly be reversed by removing non-native fishes.

Keywords: Klamath mountains, non-native ﬁsh, predator effects, subalpine lakes, subsidy

Larval insects serve as prey for larger aquatic insects, Introduction amphibians and ﬁshes, and the winged adult stages Aquatic insects are major components of both fresh- feed terrestrial predators such as birds, bats and water communities and adjacent terrestrial habitats. spiders (Power & Rainey, 2000; Nakano & Murakami,

Correspondence: Karen L. Pope, University of California, Davis, One Shields Ave., Davis, CA 95616, U.S.A. E-mail: [email protected] 1Present address: USDA Forest Service, Paciﬁc Southwest Research Station, Redwood Sciences Laboratory, 1700 Bayview Dr. Arcata, CA 95521, U.S.A.

2001; Sanzone et al., 2003). The amount of allochtho trout likely still have substantial effects on the nous insect subsidy entering a terrestrial landscape is abundance and composition of aquatic insects within affected by predation and environmental conditions and emerging from the littoral zone of these lakes. within the aquatic habitat. We tested the effects of the Similar to other lake studies, we predicted effects widespread practice of stocking non-native fishes into would be strongest for large-bodied, mobile, and mountain lakes on the amount of insect subsidy that unprotected taxa (Carlisle & Hawkins, 1998; Blumen emerges from the lakes. To make our study treatments shine, Lodge & Hodgson, 2000; Nystrom et al., 2001; realistic, we matched them to the fisheries manage Knapp et al., 2005; Venturelli & Tonn, 2005). ment practices currently being implemented in Cali Since 2000, the California Department of Fish and fornia’s mountain lakes. Game (CDFG) has reduced the number of wilderness The distribution of predatory fishes is an important lakes that it stocks by close to half to reduce impacts to determinant of the distribution and abundance of native species, especially declining amphibians such aquatic invertebrates (Brooks & Dodson, 1965; Well- as the mountain yellow-legged frog (Rana muscosa, born, Skelly & Werner, 1996; Knapp, Matthews & Camp) (Knapp & Matthews, 2000) and the Cascades Sarnelle, 2001; Baxter et al., 2004). Sport fish such as frog (R. cascadae, Slater) (Welsh, Pope & Boiano, 2006). rainbow trout (Oncorhynchus mykiss, Walbaum), and However, the benefits of stocking cessation are brook trout (Salvelinus fontinalis, Mitchill) are often top unclear because fish populations may persist in a predators in aquatic systems and feed heavily on high proportion of lakes due to local recruitment larval and emerging insects (Wellborn et al., 1996; (Armstrong & Knapp, 2004). Evidence from national Carlisle & Hawkins, 1998). These fishes have been parks in the Sierra Nevada suggests that cessation of introduced to many formerly fishless lakes and stocking is a useful long-term management option for streams throughout the world for recreation. Where restoring the aquatic invertebrate fauna (Knapp et al., trout are introduced, a proportion of the biomass that 2001, 2005), but short-term benefits of suspending formerly would have emerged and supported terres stocking are unclear (Pope, 2008). In specific lake trial life is diverted to trout (Progar & Moldenke, basins, State and Federal agencies have also begun to 2002). physically remove introduced fish to more actively Within the spatially confined habitats of small lakes, improve conditions for sensitive amphibians (Knapp, these highly mobile predators can produce strong top- Boiano & Vredenburg, 2007), but the recovery of the down effects (Vadeboncoeur et al., 2003; McCann, aquatic insect community after these top predators are Rasmussen & Umbanhowar, 2005). Most past studies abruptly eradicated has rarely been studied on the have focused on lake systems in high elevations or whole-lake scale (Finlay & Vredenburg, 2007). Addi boreal zones (Carlisle & Hawkins, 1998; Knapp et al., tional work is also needed to demonstrate that results 2001; Donald & Anderson, 2003) where predation generalise to other geographical areas. effects are expected to be strongest. In a study in We conducted a replicated whole-lake experiment Yosemite National Park, Knapp et al. (2005) found in a northern California wilderness area to test the widely varying effects of introduced trout on native effects of introduced trout at a scale relevant to lake biota: sites with a relatively low density of fish ecosystem subsidies. Study treatments were matched showed strong fish effects in alpine lakes but not at to the fisheries management practices currently being lower elevation sites. Lower elevation montane lakes implemented in California and include (i) continued often feature greater habitat complexity with differing stocking with non-native trout, (ii) suspension of substrates, more downed wood and aquatic macro stocking, and (iii) cessation of stocking and removal of phytes (Lacoul & Freedman, 2006) compared to alpine fish. In addition to these three treatments, we also lakes. included fish-free lakes as ‘reference’ sites. We com Structural complexity may ameliorate predation pared the biomass and abundance of emerging intensity by providing refugia for insect prey (Gilin aquatic insects at the water surface and the abundance sky, 1984; Diehl, 1988, 1992; Carlisle & Hawkins, 1998; of larval aquatic insects in the benthos among treat McCann et al., 2005). Although refuges could reduce ment and reference lakes. Thus, we were able to assess direct predator effects, because trout are large and the impacts of introduced trout on the aquatic insect active predators we hypothesised that introduced community and the recovery of the insects following

© 2008 The Authors, Journal compilation © 2008 Blackwell Publishing Ltd, Freshwater Biology, 54, 982–993 984 K. L. Pope et al. implementation of treatments designed to reduce 2210 m in elevation, ranged from 0.3 to 1.98 ha, and trout effects on native fauna. We also used regression were between 2.4 and 11.3 m deep (Table 1). Lakes analyses to assess the relative importance of ﬁsh and occurred within mixed conifer to sub-alpine habitat environmental factors on aquatic insect abundance. zones with common trees being red and white ﬁr, mountain hemlock, lodgepole pine and western white pine. Methods Study area Study design

The study was conducted in the Trinity Alps Wilder The 16 study basins were blocked into four groups of ness in the Klamath-Siskiyou Mountains of northern four lakes based on geographic location. Each group California (Fig. 1). No native fishes historically contained one historically fishless lake. The manipu occurred in the lentic habitats of the Trinity Alps lative treatments – continuation of stocking, suspen Wilderness, while recent surveys indicated that intro sion of stocking, and fish removal lakes – were duced trout occurred in approximately 85% of the randomly assigned to the three fish-containing lakes lakes greater than 2 m deep (Welsh et al., 2006). For in each group (Table 1). Although we tried to match the study, we selected 16 headwater lakes distributed the physical parameters of the fish-containing and throughout the eastern half of the wilderness (Fig. 1), fishless lakes, due to prior stocking of nearly all larger where the vast majority of lakes are found. All lakes lakes, the fish-free sites were smaller [0.64 ha ± 0.2 were relatively small (<2 ha), had low recreational (mean ± SE)] and shallower (2.67 m ± 0.15) than the use, moderate elevations, and had inflows and fish treatment lakes (1.17 ha ± 0.15 and 5.02 m ± 0.33, outflows with fish barriers so treatments would not respectively, Table 1). We conducted pre-treatment be affected by immigration of trout from outside sampling in June–August 2003 and initiated treat habitats. Four of the 16 selected lakes were naturally ments in September 2003. Crews removed trout from fishless and 12 supported introduced trout. The 12 the four removal lakes in fall and winter of 2003 and fish-containing lakes had been stocked by CDFG with spring of 2004 using multiple repeated gill net sets as brook and ⁄or rainbow trout for over 30 years prior to described by Knapp & Matthews (1998). The CDFG the start of the study. Lakes were between 1896 and maintained the fish treatments throughout the study

Fig. 1 Map of the Trinity Alps Wilderness study area with lakes differentiated by treatment.

Table 1 Physical parameters of the 16 study lakes

Elev. Max Area Lake Aq. Silt Lake Block Treatment (m) depth (m) (ha) pH temp. veg. Wood substrate

Eagle creek 1 Reference 1922 2.4 0.33 5.46 1127 0.60 0.15 0.92 Section Line 1 Removal 2182 4.1 0.99 6.38 1006 0.34 0.12 0.68 Mavis 1 Stocked 2042 4.2 1.52 6.88 1079 0.61 0.25 0.80 Hidden 1 Suspended 2050 4.6 1.51 5.99 1053 0.65 0.28 0.59 Found 2 Reference 2088 2.7 1.14 5.79 1060 0.41 0.07 0.60 Adams 2 Removal 1896 4.9 0.67 8.21 1059 0.70 0.19 0.96 Upper Stoddard 2 Stocked 1951 3.5 0.24 6.65 968 0.60 0.29 0.93 Lion 2 Suspended 2135 11.3 1.44 6.12 944 0.51 0.11 0.88 Shimmy 3 Reference 1958 3.0 0.78 6.42 1192 0.50 0.00 0.42 Little Caribou 3 Removal 2191 5.3 1.32 5.90 977 0.48 0.08 0.64 Ward 3 Stocked 2172 7.0 1.98 6.67 1013 0.75 0.15 0.68 Salmon 3 Suspended 2179 4.0 0.66 6.27 880 0.51 0.09 0.84 C26062 4 Reference 2056 2.4 0.31 7.26 764 0.43 0.01 0.83 Echo 4 Removal 2213 5.2 1.30 7.90 719 0.27 0.00 0.64 Deer 4 Stocked 2179 5.8 1.31 8.33 804 0.28 0.01 0.68 Luella 4 Suspended 2117 3.8 0.94 8.74 689 0.31 0.03 0.44

Lake temperature is the sum of the daily average water temperature for the survey period. Values for aquatic vegetation, wood, and fine substrate represent the proportion of littoral zone transects where those habitat features were encountered. by stocking the stocked lakes yearly with rainbow trout and withholding stocking from the stocking suspension and removal lakes. Brook trout had been stocked in the lakes prior to 2002 and all fish treatments supported brook trout populations at the beginning of the project. Two of the stocked lakes (Mavis and Upper Stoddard) were accidentally left unstocked in 2004. Although these lakes still sup ported fish, densities were very low in 2004 and 2005 until stocking occurred in mid-summer 2005. Post- treatment sampling was conducted at the 16 study basins every 2 weeks from June to September in 2004 through 2006 (start date was dependent on spring thaw). Six survey trips were conducted during both summers of 2004 and 2005 and five were conducted in Fig. 2 Crew member collecting insects from an emergence trap. 2003 and 2006.

was set in silt ⁄emergent vegetation habitat with the Biotic variables shallow side of the trap on the shoreline, the second Emergence To quantify aquatic insect emergence from trap was set in shallow (30–40 cm) water within 3 m the littoral zone of each lake over the course of the of the shoreline over the dominant lake substrate, and summer, we set three 0.6 m diameter floating emer the third trap was set near the second trap in about gence traps for approximately 40 h at each lake every 1 m deep water; substrate was not predetermined for sampling period. The lightweight, collapsible emer this location. Traps were emptied both mornings of gence traps were composed of polyester ‘No-See-um’ every sampling period by inserting an aspirator into a fabric secured to a tension frame dome made from sleeve in the trap and collecting all insects found fiberglass rods and floated by an inflated bicycle tire (Fig. 2). Samples were preserved in 70% ethanol and inner tube (Fig. 2). Traps were set on the west side of taken back to the lab, where they were identified to each lake away from established campsites. One trap family, counted and measured. For analyses of

© 2008 The Authors, Journal compilation © 2008 Blackwell Publishing Ltd, Freshwater Biology, 54, 982–993 986 K. L. Pope et al. emerging insects, we grouped the insects into four showed that CPUE and density are highly correlated categories: trichopterans, ephemeropterans, dipterans (r = 0.97, P < 0.01). No fish other than trout were and predators (Odonata, Megaloptera, Coleoptera). present in any of the study lakes. We chose these groups based on possible differences in their responses to trout. Most trichopterans have Environmental variables cases that help protect them from predators, ephem eropterans are relatively unprotected, and many We recorded water temperature at each study lake aquatic dipterans are infaunal and thus less exposed every 2 h during the survey season at 0.5 and 1.25 m to trout until emergence. Predatory insects differ from depth with Onset© (Pocasset, MA, U.S.A.) water the other groups because they may compete with temperature loggers. Lake temperatures were quan trout as well as being prey to them. tified by calculating average daily temperatures during the 57 sampling days when temperatures Benthos Littoral benthic macroinvertebrates were were collected at all lakes and then summing these sampled intensively during the 2006 field season. daily averages for each lake. We also recorded pH We collected 12, 1 m benthic sweep samples at each with handheld pH meters. Littoral zone habitat lake during all five sampling trips (one fish removal characteristics were measured by sampling from lake was not sampled during the first trip due to a approximately 25 evenly spaced transects around nearby forest fire), using a 30 cm wide D-frame dip the perimeter of each study lake. At each transect, net. At four equally spaced transects around the lake depth, substrate, and presence of aquatic vegetation shore, three sweeps were taken at approximately 0.1, and large woody debris were recorded at three 0.5 and 1.0 m deep. In the field, insects ‡4 mm were distances from shore (0.1, 0.5 and 1.0 m). Littoral sorted and identified to order except Odonata which zone substrate categories were defined by dominant was identified to suborder. Within each taxon, indi particle size (e.g. fines included silt and sand parti viduals were grouped by size classes, counted and cles and were <2 mm in diameter, gravels ranged measured. For benthic analyses we used the same from 2 to 32 mm in diameter; and boulders were categories described for the emerging insects but >256 mm in diameter). Estimated littoral zone slope separated the predator group into small predators was calculated by obtaining the slope of a least- (<12 mm) and large predators (>12 mm), because squares line through the three depths at each specific trout are visual predators that preferentially prey on distance from shore. The slopes for each transect large-bodied, active invertebrates (Diehl, 1992; Car- were averaged to obtain a mean littoral zone slope lisle & Hawkins, 1998). These size categories were not for each lake. used for emerging predators because they were all large at emergence. Because benthic insects could be Data analyses reliably identified in the field, we were able to use non-destructive sampling and returned the insects to We compared abundances of fish, emerging and the lake following processing. benthic insects and the biomass of emerging insects among treatments with analysis of variance. We first Trout In mid-summer of all four study years, we compared trout densities in the 12 treatment lakes in sampled trout presence and density at each fish 2003 to ensure that pre-treatment densities were not removal, continue stocking, and suspend stocking significantly different among treatments. We used lake. A single 36 m long, variable mesh, monofilament repeated measures ANOVAA to compare catch per net- gill-net was set perpendicular to the shoreline for hour in stocked lakes versus stocking suspension approximately 4 h during the daytime in each lake. lakes from 2004 to 2006 to see if differences occurred Captured trout were identified to species and post-treatment. Removal lakes were not included in counted. Trout densities were estimated as catch per this analysis because after 2003 we did not catch any unit effort (CPUE: number of fish captured per hour trout in the removal lakes during the 4-h gill-net of net set). A linear regression comparing summer samples. We also compared mean length of trout in 2003 CPUE values to actual density of fish removed stocked lakes versus stocking suspension lakes from from the four fish removal lakes in the fall of 2003 2004 to 2006.

To analyse the effect of the fish treatments on the abundances of the insect groups from 2006 at each biomass of emerging aquatic insects, we first con lake as dependent variables and density of trout verted insect length data to biomass following order- (catch per net hour), pH, lake temperature, and specific length-weight regressions given in Sabo, habitat complexity as continuous predictor variables. Bastow & Power (2002). A preliminary model in Predictor variables were selected a priori based on cluded 2003 biomass (prior to implementation of existing literature or field observations that suggested treatments) and lake area as covariates and average their importance in affecting aquatic insect abun emergent biomass from 2004, 2005, and 2006 as the dances. We created the habitat complexity variable by response variables. The covariates were not significant combining the variables for aquatic vegetation, large and were removed from the model for the final woody debris, and substrate in a principal compo analysis. Biomasses were log-transformed to meet nents analysis and using the canonical scores from the ANOVANOVAA assumptions. first axis, which accounted for 72% of the explained Separate MANOVA s were run on the emergence and variance. The eigenvectors for the three variables were benthic data sets with abundances from each insect positively correlated along the first axis so that lakes group as the response variables to test for differences with a high proportion of aquatic vegetation also in relation to the four treatments. When the MANOVAs tended to have a high proportion of wood and silt produced significant results or strong trends (P < 0.1), substrate. We removed littoral zone slope as a we followed with a series of ‘protected’ ANOVAA son predictor variable due to its high collinearity with each response variable to evaluate which variables lake water temperature (r = 0.82). contributed significantly to the multivariate responses To assess how the aquatic insect community related (Scheiner, 2001). We analysed the abundances of to the environmental variables (including fish den emerging insects during the 3 years of the experiment sity), we first ran a multivariate multiple regression on using repeated measures ANOVANOVAAs. The 2003 abun both the emergence and benthic data using canonical dance covariate was significant for Ephemeroptera correspondence analysis (CCA). Results of the CCAs (P = 0.005) and the lake area covariate was significant were tested for relationships between environmental for Trichoptera (P = 0.02). These covariates were not and insect community matrices with 1000 Monte significant for other groups and were removed from Carlo simulations. If significant (P £ 0.05), we fol their models. Benthic data was only collected in 2006. lowed with univariate multiple regressions to calcu We used repeated measures ANOVAA s to compare late the relative importance of predictor variables treatment effects over the course of the sampling influencing the abundances of insects in the four season, with survey period as the repeated factor and insect groups. For each group, we ran all-possible depth as an additional within-lake predictor. All subsets regressions and ranked the 15 models with abundances were log-transformed to meet normality AICc (Burnham & Anderson, 2002). To provide a assumptions. metric for the importance of each variable in the To test for the specific effect of fish removal and context of the set of variables considered, we com stocking suspension on aquatic insect biomass, we puted Akaike weights for all models within four AICc included a priori contrasts between these treatments units of the top model and then summed the weights and stocked lakes. ANOVANOVAAs were consistent with of models containing the particular variable (Burn assumptions of homoscedasticity, normality of resid ham & Anderson, 2002). The CCA analyses were uals (with the exception of emerging Ephemeroptera), performed with PCord 4.0 (MjM Software Design, and additivity between block and treatment effects. Gleneden Beach, OR, U.S.A.) and univariate regres Two-tailed P-values were used in all instances. Anal sions were performed with SAS. yses of treatment effects were conducted with PROC GLM and PROC MIXED in SAS 8.0 (SAS Institute, Carey, NC, U.S.A.). Results We used multiple regression and ordination to Trout determine the relative contributions of fish density and environmental variables to the abundance of In 2003, pre-treatment densities of trout were similar benthic and emerging aquatic insects. We used in the three fish treatment categories (F2 = 0.25,

© 2008 The Authors, Journal compilation © 2008 Blackwell Publishing Ltd, Freshwater Biology, 54, 982–993 988 K. L. Pope et al. P = 0.784). We removed 672 trout (94% S. fontinalis were odonates, primarily damselﬂies of the family and 6% O. mykiss) from the four ﬁsh removal lakes in Coenagrionidae. Aeshnidae, Libellulidae, Corduliidae the fall and winter of 2003 and did not catch trout and Lestidae were also represented as were Megalop again in the 2004–06 4-h gill net sets. Across all post- tera, family Sialidae. Emerging ephemeropterans were treatment years, stocking suspension lakes did not of the family Baetidae and trichopterans were primarily differ from stocked lakes in terms of trout densities from the families Leptoceridae, Limnephilidae and

(F1,9 = 1.31, P = 0.32) or mean sizes (F1,9 = 1.28, Sericostomatidae. Midges from the families Chironom P = 0.3), although fish densities tended to be greater idae and Ceratopogonidae made up most of the and sizes tended to be smaller in the stocked lakes dipteran group. than stocking suspension lakes (Pope, 2008). Abundance patterns were similar to biomass for emerging insects, with an increasing number of Trichoptera, Ephemeroptera, and predators emerging Emergence from the removal lakes compared to the other treat In 2006, 3 years after fish removal, fish removal lakes ment lakes (Fig. 4). By 2006, more insect predators had substantially greater emerging biomass than and Trichoptera emerged from removal lakes than stocked lakes, which had the lowest of all treatments stocked lakes (removal-stocked contrast, F1,9 = 5.97,

(Tukey’s HSD: P = 0.04). Stocking suspension lakes did P = 0.04; and F1,8 = 13.40, P = 0.006, respectively). In not differ from stocked lakes (Tukey’s HSD: P = 0.66). contrast, there was a tendency for fish removal lakes Emerging insect biomass tended to be greater at to have fewer Diptera emerging than stocked lakes removal lakes compared to stocked lakes over all (F1,9 = 3.79, P = 0.08). By 2006, removal lakes did not post-treatment years (Fig. 3, F1,9 = 4.07, P = 0.07), even significantly differ from the reference lakes for any of though two stocked lakes showed significant biomass the individual insect groups (Tukey’s HSD: P > 0.22 increases in 2005 coincident with a failure to stock those in all cases), although combined abundances of lakes in 2004 and a subsequent reduction in fish Trichoptera, Ephemeroptera, and predators at the densities. Biomass was more variable in reference and removal lakes surpassed the reference lakes (Fig. 4). stocking suspended lakes throughout the experiment, There were no differences between stocking suspen and including these lakes in the analysis weakened the sion lakes and stocked lakes for any of the insect overall difference between treatments (F3,9 = 2.67, groups (P > 0.18 in all cases). P = 0.11). Predators represented the majority of Multiple regression analyses indicated that fish emerging aquatic insect biomass caught in traps density (CPUE) was the most important predictor of (89%) and over 95% of the emerging predators abundances of emerging Ephemeroptera, Trichoptera

18 20 000 Fish removal lakes Reference 16 Stocked lakes Removal 14 Stock 15 000 12 Suspend 10 10 000 8

Biomass (g) 6 5000 4 2

0 of ephemeroptera, Mean number trichoptera, and predators, combined trichoptera, and predators, 0 2003 2004 2005 2006 Ye a r 2003 2004 2005 2006 Year Fig. 3 Mean annual biomass (±SE) of emerging insects that were caught in emergence traps each year for the ﬁsh removal lakes Fig. 4 The combined mean number (±SE) of Ephemeroptera, and stocked lakes. Two stocked lakes were accidentally not Trichoptera, and insect predators that were caught in emergence stocked in 2004 with a subsequent decrease in ﬁsh density and traps each year by treatment category. 2003 was prior to increase in emerging insect biomass in the 2005 sampling year. implementation of treatments.

Table 2 Results of all-possible-subsets regressions predicting the abundance of emerging and benthic insect groups

Variable weights No. Insect group Survey Habitat CPUE Lake temp. pH Adj. R2 of models

Ephemeroptera Emergence 0.15+ *1.0) 0.19) 0.52+ 0.50 6 Benthic 0.05+ *0.92) 0.10) 0.38+ 0.57 5 Trichoptera Emergence 0.14) *1.0) 0.11) 0.12+ 0.43 4 Benthic 0.06) *0.83) 0.45) *0.50+ 0.33 7 Diptera Emergence 0.11+ *1.0+ 0.11) 0.11) 0.46 4 Benthic 0.18+ *0.49+ 0.30) 0.19) 0.13 8 Predators Emergence 0.36+ 0.11) *0.90+ 0.09+ 0.45 5 Lg. predators Benthic *0.61+ 0.16) 0.37+ *0.88) 0.51 7 Sm. predators Benthic 0.17+ *1.0) 0.15+ 0.53+ 0.23 6

For each variable, we report the sum of AICc weights for all models within 4 AICc units of the top model in which the variable occurred. Signs indicate the relationship of the environmental variable with the dependent variable. The adjusted R2 value is for the top model based on AICc. The number of additional models within 4 AICc units of the top model is also reported. CPUE, catch per unit effort.

*Variable was in top model based on AICc. and Diptera, while lake temperature was most impor but effects were non-significant for other groups tant for predicting abundance of predators (Table 2). (P > 0.23 in all cases). Fish had a negative effect on the abundance of Although there was no indication that fish treat Ephemeroptera, Trichoptera and predators and a ment affected the overall abundance of larval insect positive effect on the abundance of Diptera. The first predators, treatments did affect their distribution two axes interpreted by CCA were significant (axis 1: within lakes. Predators ‡12 mm were more common P = 0.009, axis 2: P = 0.006) and explained 35% and in deeper water in fish removal lakes compared to 18% of the explained variance (70% of the total stocked lakes (depth by treatment interaction: variance), respectively. The axes were most strongly F3,6 = 3.43, P = 0.03). This group was dominated by correlated with fish density and lake temperature Odonata (81%), equally represented by damselflies (Fig. 5a). and dragonflies, with Megaloptera and Coleoptera constituting the remainder. The first axis of the CCA relating insect abundances Benthos to fish and environmental variables was the only The abundance of benthic insects was similarly significant axis (P = 0.03) and explained 49% of the affected by fish treatment (MAMANOVANOVA : F15,14 = 2.41, variance. The axis was most highly correlated with P = 0.05), with significant treatment effects on fish density (r = 0.70, Fig. 5b). Fish density was an

Ephemeroptera (F3,9 = 3.98, P = 0.05) and Trichoptera important negative predictor of abundances of

(F3,9 = 3.83, P = 0.05). Trichoptera were more abun Ephemeroptera, Trichoptera, and small predators dant in reference lakes than lakes in which stocking based on the univariate multiple regression models was suspended (Tukey’s HSD, P = 0.05), and the (Table 2). Large predators were most strongly influ tendency was similar for Ephemeroptera (Tukey’s enced by habitat complexity and pH. They were more HSD, P = 0.09). Neither Diptera nor large and small abundant in the benthos of complex littoral zones and predators showed differences in abundance among in lakes with moderate to low pHs. treatments (P > 0.18 in all cases). Because we only counted insects ‡4 mm, this analysis did not include Discussion most dipterans, especially the small midges that were more abundant emerging from lakes with fish. When This whole-lake, replicated experiment showed that we focused on the comparison of fish removal lakes the presence of introduced trout was the most and lakes that continued to be stocked with fish, we important factor affecting the emergence of insects found that fish removal tended to increase the from mid-elevation lakes in northern California. The abundance of Ephemeroptera (F1,9 = 4.36, P = 0.07), abundance of ephemeropterans, trichopterans and

(a) & Vredenburg, 2007; Marczak & Richardson, 2007). Nakano & Murakami (2001) showed that emerging insects constituted on average 25% of the annual energy budget for riparian birds and up to 90% for some species seasonally, while McCarty (1997) docu mented strong increases in swallow foraging rates over experimental ponds where insect emergence had increased in response to fish removal. Changes in aquatic insect emergence can have similar effects on bats (Fukui et al., 2006). When trout were removed from four lakes, overall insect biomass quickly increased compared to lakes still containing trout, and after 3 years the removal lakes surpassed the fish-free reference lakes in abun dances of large-bodied insects and biomass of emerg (b) ing insects. Removal lakes may have surpassed reference lakes because reference lakes were smaller and less complex than the treatment lakes; perhaps such habitats cannot support the same amount of insects as the larger lakes. Alternatively, a lack of alternative vertebrate predators following trout removal may have allowed aquatic insects to reach higher densities in the short-term. For example, fish are negatively correlated with some predatory salamanders in this area (e.g. Welsh et al., 2006). Additionally, the fish removal lakes may have had a buildup of nutrients (Schindler, Knapp & Leavitt, 2001) that fueled better insect breeding and survival in the recovery period. Even though most lakes had complex littoral zones Fig. 5 Ordination of sites along the first two axes of the that might have ameliorated predator effects, our Canonical Correspondence Analysis regressing (a) emerging results are consistent with the results of several and (b) benthic aquatic insect abundances on environmental studies conducted in less complex alpine or high data, using LC scores. The bi-plot overlay shows vectors related to the environmental variables. The longer the vector, the latitude habitats (Carlisle & Hawkins, 1998; Knapp stronger the relationship with the insect community matrix. The et al., 2001; Nystrom et al., 2001; Venturelli & Tonn, position of the insect groups (+) relative to the environmental 2005). In general, large-bodied insects were less vectors can be used to interpret the relationship between insect abundant in lakes with trout, while dipterans, which groups and environmental variables. Axes are scaled by stan dard deviates. were predominantly small and occurred in the sub strate, tended to be more abundant. The increase in insect predators emerging from lakes was inversely dipterans may be due to decreased competition from related to densities of trout. This is important because larger grazing or detritivorous insects or to predator changes in insect emergence can have cascading release if fish preferentially preyed on large-bodied consequences for terrestrial communities. For exam invertebrate predators (Gilinsky, 1984; Blumenshine ple, Knight et al. (2005) found that fish reduced et al., 2000; Tolonen et al., 2003). Habitat complexity dragonfly emergence with subsequent consequences was associated with greater abundance of large for terrestrial plants. The overall reduction of insect predatory insects. However other groups did not biomass due to fish is also important for terrestrial show this pattern, leaving trout as the most important predators (McCarty, 1997; Nakano & Murakami, driver of insect abundance. These results are impor 2001; Power et al., 2004; Fukui et al., 2006; Finlay tant in showing that results of earlier studies

© 2008 The Authors, Journal compilation © 2008 Blackwell Publishing Ltd, Freshwater Biology, 54, 982–993 Trout alter aquatic insect emergence 991 generalise to lower elevations and a different geo Recovery of some insect taxa apparently took as graphical area. little as 1 year in fish removal lakes, and, in general, The single year of benthic data showed similar, but insect communities in fish removal lakes did not differ not identical patterns of fish effects on large-bodied from fishless reference lakes at the end of the 3-year taxa. Ephemeroptera larvae responded positively to recovery period. However, suspension of stocking fish removal and Trichoptera larvae tended to be more was not effective in restoring insect abundance or abundant in lakes without fish, but we did not detect biomass for most of the lakes in our study. The lack of treatment effects on insect predators or dipterans. We effect was not surprising because stocking suspension did, however, find indications that the presence of did not significantly affect trout density during the trout changed the distribution of large predaceous experiment. By 2006, however, one of the stocking insects in the benthos. More large predators occurred suspension lakes appeared to have gone fishless. Four in shallow water in stocked lakes compared to fishless times as many large-bodied insects emerged from that lakes. This difference in spatial distribution could lake in 2006 compared to 2003, when fish density was result from habitat-specific predation rates, behavioral highest. The results of our regression analyses suggest changes to avoid fish, or both (Crumrine, 2006; that insect abundances respond to fish density and not Wohlfahrt et al., 2006). We did not find similar effects simply presence ⁄absence. Additional monitoring will on small predators. This is consistent with previous be necessary to reveal if and when fish densities studies showing that the behavioral avoidance res decrease enough in the stocking suspension lakes to ponse of older instars is greater than that of young see significant changes in aquatic insect populations. instars (Crumrine, 2006). The lack of significant In conclusion, we were able to assess the short-term differences across treatments with regard to abun ramifications of changes in fisheries management dance of large predators seemed to be influenced by practices on aquatic insect subsidies. We showed habitat complexity, pH and water temperature. Our rapid recovery of large-bodied insect taxa in mid- multiple regression analyses showed that predator elevation lakes that were restored to a fishless condi abundances were more influenced by these factors tion, whereas recovery was slower or absent in lakes than by trout. We believe that we inadvertently where stocking was suspended. The vast majority of sampled across the altitudinal limits of Odonates mountain lakes in the western United States have because they were far less abundant at higher altitude been stocked with trout for several decades. Impacts lakes than at lower altitude lakes. Thus, environmental of this large-scale fish stocking effort on aquatic-to factors superseded fish treatment effects for the large terrestrial ecosystem subsidy have just begun to be predator group. Given our low replication and high assessed (Finlay & Vredenburg, 2007), and insights variability across lakes, we likely lacked power to from in-depth studies following lake restoration to a elucidate the more subtle effects of fish on large insect fishless condition will help scientists and wildlife predators in the benthos of complex littoral zones. managers understand the impacts of past actions and Overall, effects on benthos tended to be slightly the consequences of future management changes. weaker compared to emergence, which could be due to the more limited time over which samples were Acknowledgments collected or because benthic samples collected early in the season may have contained larvae that would We thank all the excellent field biologists who helped have been eliminated by predation later in the season. collect the data. H. Welsh, J. Garwood and W. Rainey Many taxa would be especially vulnerable to fish provided insight and expertise throughout the project. predators as they leave structural refuges (i.e. caddis G. Hodgson and N. Willits helped with statistical fly cases, detritus) or simply move within the aquatic analyses. R. Knapp, T. Schoener, P. Moyle, G. Gross habitat to accomplish emergence. Alternatively, fish man and two anonymous reviewers offered helpful effects may have been equally strong in the benthos, comments to improve this paper. Thanks to CDFG, but noise in the data due to sampling issues may have especially B. Bolster, B. Aguilar, and E. Pert for their contributed to the reduced effects. Sweep sampling is endorsement and logistical support of the project. difficult in habitats with woody debris or dense This research was funded by CDFG (ESA Section 6 aquatic vegetation. grants E-2-F-21 and E-2-F-27), the National Science

© 2008 The Authors, Journal compilation © 2008 Blackwell Publishing Ltd, Freshwater Biology, 54, 982–993 992 K. L. Pope et al. Foundation (DEB 0415505), UC Water Resources Knapp R.A. & Matthews K.R. (1998) Eradication of Center (W-987), and the UC Davis Wildlife Health nonnative fish by gill netting from a small mountain Center. KLP received additional support from the US lake in California. Restoration Ecology, 6, 207–213. Forest Service Pacific Southwest Research Station and Knapp R.A. & Matthews K.R. (2000) Non-native fish UC Davis Ecology Block Grants. JP-S received support introductions and the decline of the mountain yellow- legged frog from within protected areas. Conservation from the UC Davis Center for Population Biology. Biology, 14, 428–438. Knapp R.A., Matthews K.R. & Sarnelle O. (2001) Resis References tance and resilience of alpine lake fauna to fish introductions. Ecological Monographs, 71, 401–421. Armstrong T.W. & Knapp R.A. (2004) Response by trout Knapp R.A., Hawkins C.P., Ladau J. & McClory J.G. populations in alpine lakes to an experimental halt to (2005) Fauna of Yosemite National Park lakes has low stocking. Canadian Journal of Fisheries and Aquatic resistance but high resilience to fish introductions. Sciences, 61, 2025–2037. Ecological Applications, 15, 835–847. Baxter C.V., Fausch K.D., Murakami M. & Chapman P.L. Knapp R.A., Boiano D.M. & Vredenburg V.T. (2007) (2004) Fish invasion restructures stream and forest Removal of nonnative fish results in population food webs by interrupting reciprocal prey subsidies. expansion of a declining amphibian (mountain yel Ecology, 85, 2656–2663. low-legged frog, Rana muscosa). Biological Conservation, Blumenshine S.C., Lodge D.M. & Hodgson J.R. (2000) 135, 11–20. Gradient of fish predation alters body size distribu Knight T.M., McCoy M.W., Chase J.M., McCoy K.A. & tions of lake benthos. Ecology, 81, 374–386. Holt R.D. (2005) Trophic cascades across ecosystems. Brooks J.L. & Dodson S.I. (1965) Predation, body size, Nature, 437, 880–883. and composition of plankton. Science, 150, 28–35. Lacoul P. & Freedman B. (2006) Relationships between Burnham K.P. & Anderson D.R. (2002) Model Selection aquatic plants and environmental factors along a steep and Multimodel Inference, 2nd edn. Springer, New York. Himalayan altitudinal gradient. Aquatic Botany 84, Carlisle D.M. & Hawkins C.P. (1998) Relationships 3–16. between invertebrate assemblage structure, two trout Marczak L.B. & Richardson J.S. (2007) Spiders and species, and habitat structure in Utah mountain lakes. subsidies: results from the riparian zone of a coastal Journal of the North American Benthological Society, 17, temperate rainforest. Journal of Animal Ecology, 76, 286–300. 687–694. Crumrine P.W. (2006) Age specific behavioral responses McCann K.S., Rasmussen J.B. & Umbanhowar J. (2005) of odonate larvae to chemical and visual cues from The dynamics of spatially coupled food webs. Ecology predators. Journal of Freshwater Ecology, 21, 9–16. Letters, 8, 513–523. Diehl S. (1988) Foraging efficiency of three freshwater McCarty J.P. (1997) Aquatic community characteristics fishes: effects of structural complexity and light. Oikos, influence the foraging patterns of tree swallows. 53, 207–214. Condor, 99, 210–213. Diehl S. (1992) Fish predation and benthic community Nakano S. & Murakami M. (2001) Reciprocal subsidies: structure: the role of omnivory and habitat complexity. dynamic interdependence between terrestrial and Ecology, 73, 1646–1661. aquatic food webs. Proceedings of the National Acad Donald D.B. & Anderson R.S. (2003) Resistance of the emy of Sciences of the United States of America, 98, prey-to-predator ratio to environmental gradients and 166–170. to biomanipulations. Ecology, 84, 2387–2394. Nystrom P., Svensson O., Lardner B., Bronmark C. & Finlay J.C. & Vredenburg V.T. (2007) Introduced trout Graneli W. (2001) The influence of multiple introduced sever trophic connections in watersheds: predators on a littoral pond community. Ecology, 82, consequences for a declining amphibian. Ecology, 1023–1039. 88, 2187–2198. Pope K.L. (2008) Assessing changes in amphibian Fukui D., Murakami M., Nakano S. & Aoi T. (2006) population dynamics following experimental mani Effect of emergent aquatic insects on bat foraging in pulations of introduced fish. Conservation Biology a riparian forest. Journal of Animal Ecology, 75, (in press). 1252–1258. Power M.E. & Rainey W.E. (2000) Food webs and Gilinsky E. (1984) The role of fish predation and spatial resource sheds: towards spatially delimiting trophic heterogeneity in determining benthic community interactions. In: The Ecological Consequences of Environ structure. Ecology, 65, 455–468.

mental Heterogeneity (eds M.J. Hutchings, M.J. John & Tolonen K.T., Hamalainen H., Holopainen I.J., Mikkonen A.J.A. Stewart ) pp. 291–314. Blackwell Science, Mal- K. & Karjalainen J. (2003) Body size and substrate den. association of littoral insects in relation to vegetation Power M.E., Rainey W.E., Parker M.S., Sabo J.L., Smyth structure. Hydrobiologia, 499, 179–190. A., Khandwala S., Finlay J.C., McNeely F.C., Marsee K. Vadeboncoeur Y., Jeppesen E., Vander Zanden M.J., & Anderson C. (2004) River-to-watershed subsidies in Schierup H.H., Christoffersen K. & Lodge D.M. (2003) an old-growth conifer forest. In: Food Webs at the From Greenland to green lakes: cultural eutrophication Landscape Level (Eds G.A. Polis, M.E. Power & G.R. and the loss of benthic pathways in lakes. Limnology Huxel), pp. 217–240. The University of Chicago Press, and Oceanography, 48, 1408–1418. Chicago. Venturelli P.A. & Tonn W.M. (2005) Invertivory by Progar R.A. & Moldenke A.R. (2002) Insect production northern pike (Esox lucius) structures communities of from temporary and perennially flowing headwater littoral macroinvertebrates in small boreal lakes. Jour streams in western Oregon. Journal of Freshwater nal of the North American Benthological Society, 24, 904– Ecology, 17, 391–407. 918. Sabo J.L., Bastow J.L. & Power M.E. (2002) Length-mass Wellborn G.A., Skelly D.K. & Werner E.E. (1996) Mech relationships for adult aquatic and terrestrial inverte anisms creating community structure across a fresh brates in a California watershed. Journal of the North water habitat gradient. Annual Review of Ecology and American Benthological Society, 21, 336–343. Systematics, 27, 337–363. Sanzone D.M., Meyer J.L., Marti E., Gardiner E.P., Tank Welsh H.H., Pope K.L. & Boiano D. (2006) Sub-alpine J.L. & Grimm N.B. (2003) Carbon and nitrogen transfer amphibian distributions related to species palatability from a desert stream to riparian predators. Oecologia, to non-native salmonids in the Klamath Mountains of 134, 238–250. northern California. Diversity and Distributions, 12, 298– Scheiner S.M. (2001) MANOVA: multiple response 309. variables and multispecies interactions. In: Design and Wohlfahrt B., Mikolajewski D.J., Joop G. & Suhling F. Analysis of Ecological Experiments (Eds S.M. Scheiner & (2006) Are behavioural traits in prey sensitive to the J. Gurevitch), pp. 99–115. Oxford University Press, risk imposed by predatory fish? Freshwater Biology, 51, New York. 76–84. Schindler D.E., Knapp R.A. & Leavitt P.R. (2001) Alter ation of nutrient cycles and algal production resulting from fish introductions in mountain lakes. Ecosystems, (Manuscript accepted 12 October 2008) 4, 308–321.